Inaccurate Lucene whitespace tokenization of soft returns

When content is separated by a soft return, Lucene doesn't see a soft return as whitespace or word boundary and gets rid of the soft return character. Impact:

This is line 106[soft return]
Next line

Gets parsed and indexed as 106Next

So searching for 106 returns nothing but searching for 106Next returns accurate results. However, obviously this isn't accurate.

Rated as Major because makes index entries flawed